The BFD for OSPF feature gives OSPF routing protocol the ability to
utilize BFD's fast failure detection to monitor OSPF neighbor adjacencies. CLI commands are
provided to configure BFD protection for OSPF, so that as a registered BFD client, OSPF can
request BFD protection for interested OSPF neighbors, and receive notifications about BFD session
setup status and BFD session status updates (after establishing) from the BFD server. When BFD
detects a communication failure between neighbors, it informs OSPF, which causes the OSPF
neighbor state be marked as "down." This allows OSPF protocol to quickly begin network
convergence and use alternate paths to the affected neighbor.
Establishing a BFD Session for OSPF Neighbor
OSPF discovers a neighbor.
If BFD for OSPF is configured, OSPF on both routers sends a request to the local BFD server
to initiate a BFD neighbor session with the OSPF neighbor router.
The BFD neighbor session with the OSPF neighbor router is established on both sides if BFD
session limit is not reached.
If the BFD session limit is reached, the OSPF neighbor will be marked as BFD session failed if
synchronous request is used, or pending if asynchronous request is used, and the BFD server will
send an asynchronous notification when the session registration passes later. (The asynchronous
request is not available until the BFD client session create API is enhanced.)
Eliminating the OSPF Neighbor Relationship by BFD Fault Detection
A failure occurs in the network.
The BFD neighbor session with the OSPF neighbor router is removed
because the BFD timer expired.
On both Router A and Router B, BFD notifies the local OSPF process that
the BFD neighbor is DOWN.
The local OSPF process tears down the OSPF neighbor relationship by
marking neighbor state DOWN. (If an alternative path is available, the routers will immediately
start converging on it.)
OSPF will request the BFD server to delete the BFD session for OSPF
neighbor moved to DOWN state. When the link failure is later resolved, OSPF needs to register
the re-discovered neighbor to BFD again to initiate BFD session creation.
Removing BFD Protection for OSPF
BFD protection is removed from OSPF interface on Router A, OSPF process
requests BFD server to delete all BFD sessions for neighbors learned on that interface.
BFD server will stop sending session status update to local OSPF process
on Router A.
Before actually deleting any sessions, BFD server on Router A will first
notify Router B to mark those session status as "Admin Down," which will cause Router B to stop
using BFD protection for those OSPF neighbors.
When BFD protection is removed from OSPF interface on Router B, BFD
sessions can be deleted immediately since they are already in "Admin Down" state.
When BFD for OSPF is configured on broadcast interface, the default behavior
is to register only OSPF neighbors in FULL state with the BFD server. Separate BFD sessions are
created for each neighbor learned on the same interface. If multiple clients ask for the same
neighbor on the same interface, then single BFD sessions are established between the peers.
OSPF Neighbor State Determination
With active BFD protection, OSPF combines the BFD session state with the
associated interface admin and operational states to determine the OSPF neighbor adjacency
discovered on that OSPF interface. Regarding OSPF neighbor relationships, OSPF reacts directly
to BFD session state changes only in the following circumstances:
If BFD is enabled on the interface, and
If BFD for OSPF is configured on the OSPF interface, and
If a BFD session has been established to the neighbor, and
If the BFD session has passed the INIT_COMPLETE state then:
The OSPF neighbor relationship will remain as FULL if the operational state
of the BFD session is "UP" and the operational state of the associated VLAN interface is UP.
The OSPF neighbor relationship will be considered as DOWN if the
operational state of the BFD session is DOWN or the operational state of the associated VLAN
interface is DOWN.
In all other cases, the BFD session state is not considered as part of the
reported OSPF neighbor state, and the OSPF neighbor state reverts to the operational state of
the OSPF interface only. When the BFD session is in ADMIN_DOWN state, OSPF ignores BFD events
and OSPF neighbor adjacency is not be affected by the BFD session state change.